(342l) Application of BONUS Algorithm for Optimal Spatio-Temporal Sensor Placement Under Uncertainty: A Real World Case Study of City of Atlanta

Rajib Mukherjee^1,2, Urmila M. Diwekar^1*, Naresh Kumar³

¹Center for Uncertain Systems: Tools for Optimization & Management, Vishwamitra Research Institute, Crystal Lake, IL 60012

²Department of Chemical Engineering, University of Texas Permian Basin, TX 79762

³Electric Power Research Institute, Palo Alto, CA 94304

Abstract

Air pollution exposure assessment involves monitoring of pollutant species concentrations in the atmosphere along with its health impact assessment on the population. Often air pollutants are monitored via stationary monitoring stations. Due to the high cost of sensors as well as maintenance of a monitoring network, sensors can only be installed at a limited number of locations [1,2]. The sparse spatial coverage of immobile monitors can lead to errors in estimating the actual exposure of pollutants. One approach to address these limitations is dynamic sensing, a new monitoring technique that adjusts the locations of portable sensors in real-time to measure the dynamic changes in air quality [3,4]. The key challenge in dynamic sensing is to develop algorithms to identify the optimal sensor locations in real-time in the face of inherent uncertainties in emissions estimates and the fate and transport of air pollutants [5]. In this work, we will present an algorithmic framework to address the challenge of sensor placement in real-time given those uncertainties. Uncertainty in the system includes location and amount of pollutants as well as meteorology leading to a stochastic optimization problem. We use the novel Better Optimization of Nonlinear Uncertain Systems (BONUS) algorithm to solve these problems [6]. Fisher information (FI) is used as the objective for optimization. We demonstrate the capability of our novel algorithm using a case study in Atlanta, Georgia [7]. Our real-time sensor placement algorithm allows, for the first time, determination of the optimal location of sensors under the spatial-temporal variability of pollutants, which cannot be accomplished by a stationary monitoring station. We present the dynamic locations of sensors for observing concentrations of pollutants as well as for observing the impacts of these pollutants on populations.

Keywords: spatiotemporal sensor placement, BONUS algorithm, weather uncertainties, stochastic optimization, exposure assessment

1. Shastri, Y., and Diwekar, U., (2006), "Sensor placement in water networks: A stochastic programming approach." Journal of water resources planning and management 132(3), 192-203.
2. Mukherjee, R., Diwekar, U. M. and Vaseashta, A., (2017), "Optimal sensor placement with mitigation strategy for water network systems under uncertainty." Computers & Chemical Engineering 103, 91-102.
3. de Almeida Oliveira, T., and Godoy, E. P., (2016), "Zigbee wireless dynamic sensor networks: feasibility analysis and implementation guide." IEEE Sensors Journal 16(11), 4614-4621.
4. Diwekar, U., and Mukherjee, R., (2017), "Optimizing spatiotemporal sensors placement for nutrient monitoring: a stochastic optimization framework." Clean Technologies and Environmental Policy 19(9), 2305-2316.
5. Wang, Z., Lou, H., Mukherjee, R., Diwekar, U. M., Olsen, T., Yazdanpanah, N., (2020), A New Area of Utilizing Industrial Internet of Things in Environmental Monitoring, Submitted
6. Sahin, K. and Diwekar, U., (2004), “Better Optimization of Nonlinear Uncertain Systems (BONUS): A New Algorithm for Stochastic Programming Using Reweighting through Kernel Density Estimation.” Annals of Operations Research , 132, 47-68
7. Mukherjee, R., Diwekar, U., Kumar, N., (2019), "Stochastic Optimization for Real-time Spatiotemporal Sensor Placement to Monitor Air Pollutants for Health Impact Assessment", Submitted